In order to improve the global search ability of the dynamic multi-swarm PSO(DMS-PSO) algorithm, the cuckoo search(CS) algorithm is introduced into DMS-PSO algorithm, the algorithm named DMS-PSO-CS is proposed, which is a bi-layer optimal algorithm. The whole swarm is dynamically divided into several small populations by using the median clustering algorithm, and each small population is optimized by using PSO algorithm in the bottom layer. The best particle from each small population is selected as a member of the swarm in the top layer, then the top swarm is deeply optimized by using CS algorithm. The DMS-PSO-CS algorithm is employed to solve the CEC 2014 test functions. In comparision with the CS algorithm and other improved PSO algorithms, the experimental results show that the proposed algorithm can obviously improve the global search ability and computation efficiency.